AINewsnow

Stable FP4 Pretraining with Transpose-Invariant 2D Block Scaling

This story is from 2026-09-03. It is preserved in the archive; the latest stories are on the live feed.

Stable FP4 Pretraining with Transpose-Invariant 2D Block Scaling Billing Support — September 03, 2026 A tensor value can receive one scale during the forward pass and a different scale after the same tensor is transposed for backpropagation. At FP4 precision, that mismatch is not a minor rounding d…

Read the full story at DEV Community — Machine Learning ↗

Timeline · 1 report

  1. 2026-09-03 16:07 · DEV Community — Machine Learning
    Stable FP4 Pretraining with Transpose-Invariant 2D Block Scaling

More stories

  1. Introducing Kimi K3 on Amazon Bedrock — AWS Machine Learning Blog
  2. Introducing Amazon SageMaker HyperPod Inference Gateway — AWS Machine Learning Blog
  3. Anthropic, OpenAI, SpaceXAI, Google sued over call to ‘pace’ AI development — Politico Technology
  4. Gemini Hacked Three Companies in First Known Breakout by Google’s AI — Wall Street Journal Technology
  5. NVIDIA CEO Jensen Huang rejects ‘AI will end the world’ claim, yet cautions ‘we should go as fast as we can but...’ — Mint AI
  6. Alibaba's Qwen3.8-Omni-Flash Cuts Video AI Costs by 89% With Agent Tool Use — AlphaSignal
  7. AI hallucination of Chinese nuclear components almost led to US military attack — Ars Technica AI
  8. The new AgentCore runtime: Elastic, optimized, and consistently fast starts — AWS Machine Learning Blog

Get the daily brief of stories like this at 6:30 every morning →